Research On Wild Bird Monitoring Based On Sound Recognitio

Bird resources are abundant in our country,and birds are friends of human beings and members of the ecological environment.Birds are able to sense small changes in their environment sensitively and are the most ideal ecological species for monitoring environmental changes.The monitoring of birds in the wild has always been the focus of ecological and environmental research.The existing monitoring methods are radar,ring station,which is more time-consuming and labor-intensive.With the development of sound recognition technology,considering that bird song and bird morphological characteristics are important biological characteristics,each bird sound is unique and easier to capture for morphological characteristics.Bird monitoring can be done by collecting and identifying bird sounds in the wild.In this paper,a bird monitoring system is designed for the needs of bird voice acquisition and recognition in the wild.The main contents of the work are as follows:(1)Noise reduction of birdsIn the process of collecting bird sounds in the wild,there will be a lot of environmental noise,and it is necessary to remove the noise to obtain pure bird sounds to improve the recognition accuracy.The traditional spectral subtraction has good denoising effect and strong real-time performance,but it will produce noise residual.This paper combines endpoint detection to update the noise spectrum in real time to improve spectral subtraction and achieve better denoising effect.(2)Extraction of bird sound feature parametersThe characteristics of bird sound recognition are studied,and the commonly used characteristic parameter Mel Cepstral Coefficient(MFCC)is introduced in detail.From the structure of the Mel filter bank used in extracting the MFCC,the MFCC has a good ability to characterize the low-frequency part of the sound signal,but is weak to the high-frequency part.By introducing the inverted Mel cepstral coefficient(IMFCC),the feature fusion of MFCC and IMFCC is carried out according to the Fisher criterion,and a new feature is obtained by taking the longer of the two.MFCC-IMFCC can more comprehensively characterize bird sound signal information and improve the accuracy of bird sound recognition.Rate.(3)Construction of bird recognition modelThe classification model is studied,and the classical machine learning algorithm Support Vector Machine(SVM)is introduced in detail.When SVM trains the model,the penalty factor C and kernel function g used will affect the classification effect and generalization ability of the model.The SVM is optimized by genetic algorithm(GA)to improve the classification effect and generalization ability of the classification model.(4)Hardware system constructionThe hardware requirements of the wild bird sound collection and recognition system are analyzed,and an embedded system with STM32H743IIT6 as the core is built.The sound acquisition is realized through the audio chip WM8978,the sound signal is preprocessed with the help of the powerful computing power of STM32H743IIT6,and the SVM classification function of the CMSIS-DSP library is called to realize bird sound recognition.